Progress 09/01/12 to 08/31/16
Outputs
Target Audience:Genetics research and agricultural community. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?A graduate student in Agricultural Engineering was trained in a multidisciplinary research team that included plant biologists. Three interns and one graduate student received training on the use of Li-COR to be able to collect good quality data and on the interpretation of physiological parameters collected with the equipment. An undergraduate student was trained when assisting with the robot operation and optimization. 10+ undergraduate students received training when assisting with field and robot operation. How have the results been disseminated to communities of interest?An oral presentation of the phenotypic data acquisition system and preliminary plant 3D reconstruction results were given at ASABE annual meeting in July 2014 by Tang's group. Salas Fernandez M.G. 2014. Genetic characterization of sorghum biomass yield components over time using field-based high-throughput phenotyping technology. ASA-CSSA-SSSA International Annual meeting, Long Beach, CA. Abstract ID 86748. Invited presentation on November 4, 2014. Salas Fernandez M.G. 2014. Field high-throughput phenotyping technologies for biomass sorghum: current status and future perspectives. R.F. Baker Plant Breeding Symposium, Iowa State University, Ames, IA. March 6, 2014. Salas Fernandez, M.G. Breeding, genomics and high-throughput phenotyping strategies to design a high-yielding biomass sorghum for biofuel production. Energy Biosciences Institute, University of Illinois, Urbana-Champaign, IL. November 2013. The parallelized stereo matching algorithm and its performance and accuracy on standard benchmark as well as our sorghum images were presented at ASABE annual meeting in July 2015. Salas Fernandez M.G. 2015. NAPB Early Career Awardee presentation. Genomics, physiology and high-throughput phenotyping strategies to improve sorghum for biofuel production. National Association of Plant Breeders Annual Meeting. Pullman, WA. July 29, 2015. Salas Fernandez M.G., Bao Y., Tang L., Schnable P. 2016. A field-based high-throughput phenotyping platform to discover the genetic architecture of sorghum biomass yield components over time. XXIV Plant and Animal Genome conference, San Diego, CA. Abstract W856. Salas Fernandez M.G. 2016. Genomics, physiology and high-throughput phenotyping to improve yield potential in bioenergy sorghum. Cornell University Plant Breeding seminar series, Ithaca, NY. September 27, 2016. Bao, Y., Tang, L., Schnable, P. S., & Fernandez, M. G. S. (2016). Infield Biomass Sorghum Yield Component Traits Extraction Pipeline Using Stereo Vision. In 2016 ASABE Annual International Meeting (p. 1). American Society of Agricultural and Biological Engineers. Bao, Y., Tang, L. (2016). Field-based Robotic Phenotyping for Sorghum Biomass Yield Component Traits Characterization Using Stereo Vision. In 5th IFAC Conference on Sensing, Control and Automation for Agriculture. Schnable has described the project in 20+seminars presented around the world over project period and below are a few most recent ones: North American Plant Phenotyping Network, Purdue University, 29 August 2016 "Field-based predictive phenomics" KDD2016 Workshop on Data Science for Food Energy and Water, San Francisco CA, 14 August 2016 "Predictive Phenomics of Plants" (Keynote talk) 17th European Congress of Biotechnology, Krakow, POLAND, 3 July 2016 "Predictive Phenomics of Plants" (Keynote talk) University of Texas A&M, College Station TX, 19 May 2016 "Predictive Phenomics of Plants" University of Georgia, Athens GA, 17 February 2016 "Predictive Plant Phenomics" What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
Impact Statement The C4 grass sorghum is one of the most promising and productive species for biofuel production in the US. One of the key breeding objectives for biomass crops is increased yield. Although total biomass yield is a function of growth rate and growth duration, growth rate is typically not constant throughout the growing season. Hence, the potential exists to identify distinct genetic loci that control growth rates and plant architecture traits at different times in the growing season. To be able to conduct such research a robust automated phentoyping system is needed. But to date most automated phenotyping systems have been laboratory- or greenhouse-based. We generated genetic marker data and analysis algorithms for identifying genetic control of growth rates in sorghum and identified a couple of genetic controls that affect yield. We also developed a field-based automated phenotyping system and associated image analysis method that has been optimized and shown to be efficient and accurate. Both researchers and breeders will benefit from our accomplishment via improved lines to generate high yielding biomass sorghum hybrids and utilizing the automated phenotyping system to obtain mass amount of phenotype data. These methods conceivably can be extended to the study and improvement of other crop species. Objective 1: Identify the genetic control of growth rates, photosynthetic rates and amounts of photo-protection, and dynamic changes in plant architecture GWAS will be performed on two sets of sorghum lines: the "Diversity Panel" and the "Yu Panel". In order to conduct GWAS study with high density, low-missing data SNPs across 291 sorghum Photoperiod-Sensitive (PS) lines, 289 lines of sorghum from the Photoperiod-Sensitive panel were multiplexed and sequenced using tunable Genotyping-by-Sequencing (tGBS) technologies. In total we generated 654 million single-end raw reads from four lanes of an Illumina Hi-seq 2000 sequencer. 457 million were retained after de-barcoding. We discovered 525,911 polymorphic sites. However, the overall missing data was unacceptably high because 180/289 lines of samples did not have a sufficient number of reads (>1 million read/line) for SNP discovery. In second attempt we conducted tGBS using Ion Torrent ProtonTM sequencing platform. About 79 Gb base pair (583 million reads) of sequencing data were generated and about 831 thousand SNPs were discovered from this panel. We also identified about 17 thousand low-missing data rate (LMD<50) SNPs. The SNPs discovered by tGBS were combined with SNPs identified from traditional Genotyping-by-Sequencing generated by Dr. Ed Buckler's and Dr. Jianming Yu's groups to conduct GWAS to identify genetic architecture of sorghum biomass yield-related plant architecture traits. Both the photoperiod sensitive and insensitive sorghum panels were planted at two experimental stations in the Ames area, IA in 2014 and 2015. Two different plot designs were implemented in each replication to collect images from very different canopy density situations. The following phenotype data were collected. Photosynthesis and light-adapted fluorescence parameters have been statistically analyzed to determine Best Linear Unbiased Predictors of genotypes for further association analysis. Weather data collected at experimental farms have been included in linear models as covariates. Plant architecture parameters including plant height, leaf number, leaf angle and stem diameter were manually collected from 30 lines and these data were utilized to develop and validate our image analysis algorithms. Stand counts, to determine the number of plants in each plot on both locations. These data will be utilized to correct our biomass yield estimates by variation in plant density. Biomass data for individual plants of the plot design were collected for a subset of 100 lines in one location in 2014. The goal was to estimate the proportion of the total aboveground biomass determined by grain. Biomass data at the plot level in both locations utilizing an experimental forage chopper in both years. These data will be utilized to predict biomass yield from digital images and to correlate with growth curves and individual plant architecture parameters. We performed correlation analysis to validate our algorithmically derived plant height and stem diameter data with ground-truth, manually collected data for a subset of lines. We completed the first genome-wide association study (GWAS) component of the project using 315 photoperiod insensitive lines and features extracted from end-of-the-season images collected in two locations in 2014. A public dataset of 136,000 GBS SNPs were utilized and a comparative analysis performed between the two algorithms developed for each test, and manual data collected and published by Dr. Salas Fernandez's group in 2016. The comparative GWAS was performed on the following phenotypic data sets: For plant height: Image-derived using a semi-automatic Dense Stereo method Image-derived using an automatic hedge-based extraction pipeline Manually collected and published in Zhao et al. (2016) For stem thickness: Image-derived stem diameter using a semi-automatic Dense Stereo method Image-derived stem diameter using semi-automatic Image Patch Stereo Matching algorithm Manually collected stem circumference and published in Zhao et al. (2016) GWAS results confirmed that our high-throughput phenotyping technology and algorithms were very accurate to estimate plant height, since the same associated regions and SNPs were identified on chromosomes 6 and 9. Even though stem diameter is a highly complex trait determined by many loci with small effects, we demonstrated that the Image Patch Stereo Matching algorithm generated robust estimations of the phenotype supported by the discovery of consistent SNPs and regions on chromosomes 1 and 7. The high correlations between ground-truth data and stem diameter estimates extracted from our stereo images (r=0.929) of the same subset of individuals provided another evidence of the quality of the algorithmically derived data. Objective 2: Develop a high-throughput phenotyping robot to measure growth rates. Cameras were calibrated to rectify image distortion and acquire relative positions and orientations. A variety of stereo matching algorithms from Middlebury stereo vision benchmark were tested. The efficiency of the data acquisition software was improved. Sorghum images were collected from the two locations alternately once per week (as weather permitted) from mid-July to the beginning of September 2014. A total about 5,131 images were collected per location per week. Image data were collected from a subset of sorghum varieties at harvest in 2014 with upgraded NIR-converted stereo cameras. We also collected NIR stereo images of two locations between mid-June to mid-August in 2015. We added Kinect sensors to current system and collected 3D point clouds of a subset of sorghum varieties in September 2015. We calibrated the NIR-converted stereo cameras to rectify image distortion and acquire their relative positions and orientations. We developed sorghum plant segmentation algorithm based on NIR images. We also developed plant localization and stalk width detection algorithm for sorghum at the early stage based on the NIR segmentation images. We developed an user-interactive software to measure stem diameter in 3D space based on the stereo image data. The results were highly correlated (0.958) to the ground truth measured with caliper in field. Based on the stereo 3D reconstruction, we developed an automated image processing pipeline to extract plot-based features including plot height, plot width, vegetation volume index and vegetation area index. The processing pipeline was able to process all the image data of entire field in about 5 hours. The phenotyping could potentially be on a daily basis.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2014
Citation:
Salas Fernandez M.G. 2014. Genetic characterization of sorghum biomass yield components over time using field-based high-throughput phenotyping technology. ASA-CSSA-SSSA International Annual meeting, Long Beach, CA. Abstract ID 86748. Invited presentation on November 4, 2014.
- Type:
Journal Articles
Status:
Other
Year Published:
2017
Citation:
Salas Fernandez M.G., Bao Y, Tang L., Schnable P. 2016. High-throughput field-based phenotyping technology for high biomass producing crops. Plant Physiology (in preparation).
|
Progress 09/01/14 to 08/31/15
Outputs
Target Audience:Genetics research and agricultural communities. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?Two interns and one graduate student received training on the use of Li-COR to be able to collect good quality data and on the interpretation of physiological parameters collected with the equipment. An undergraduate student was trained when assisting with the robot operation and optimization. 10+ undergraduate students received training while assisting with field and robot operation. How have the results been disseminated to communities of interest?Salas Fernandez M.G. 2014. Genetic characterization of sorghum biomass yield components over time using field-based high-throughput phenotyping technology. ASA-CSSA-SSSA International Annual meeting, Long Beach, CA. Abstract ID 86748. Invited presentation on November 4, 2014. Salas Fernandez M.G. 2015. NAPB Early Career Awardee presentation. Genomics, physiology and high-throughput phenotyping strategies to improve sorghum for biofuel production. National Association of Plant Breeders Annual Meeting. Pullman, WA. July 29, 2015. The parallelized stereo matching algorithm and its performance and accuracy on standard benchmark as well as our sorghum images were presented at ASABE annual meeting, New Orleans, Louisiana, July, 2015. What do you plan to do during the next reporting period to accomplish the goals? Photosynthesis and fluorescence parameters will be used in association analysis as well as all architecture parameters extracted from images. Growth curves determined from algorithmically determined parameters will be tested as predictors of biomass. Collect final biomass image data in 2015. Improve current stereo matching algorithm on robustness and speed. Develop sorghum stalk detection algorithm for all growth stages. Extract other traits such as plant height, volume and panicle size. Model and predict sorghum biomass yield based on phenotypic data. Process data collected by Kinect ToF sensor. Compare NIR stereo imaging and Kinect ToF imaging as phenotyping sensor.
Impacts
What was accomplished under these goals?
Sorghum has great potential as a biomass crop for energy production. We are applying a systematic genetic approach to study the control of biomass yield components in sorghum. Specifically, we are seeking to identify the genetic control of rates of photosynthesis, photo-protection (resistance to harmful amounts of sunlight), biomass growth rates and the dynamic changes in plant structure (such as plant height, stalk diameter, leaf number, leaf width, leaf length, leaf angle, leaf area index, etc.). Using advanced statistical approaches, we are identifying "genetic markers" that are associated with these yield components so they can be tracked within sorghum biomass breeding programs. To identify these relationships between genetic markers and biomass yield components it is necessary to collect trait data at multiple times during multiple growing seasons. It would be extremely challenging to do so using conventional approaches. Instead these data are being collected using a sophisticated, high-throughput, field-based, plant phenotyping system that is being developed as part of the project. This field-based phenotyping system has the potential to revolutionize the collection of phenotypic data from biomass yield trials. As such, this robotic system is expected contribute widely to the genetic improvement of biomass crops. We have identified the molecular markers (SNPs), collected/analyzed trait data, and developed and optimized the phenotyping robotic system. Objective 1: Identify the genetic control of growth rates, photosynthetic rates and amounts of photo-protection, and dynamic changes in plant architecture GWAS will be performed on two sets of sorghum lines: the "Diversity Panel" and the "Yu Panel". In order to conduct GWAS study with high density, low-missing data SNPs across 291 sorghum PS lines, we applied a novel tunable Genotyping-by-SequencingTM technology (tGBS) using Ion Torrent ProtonTM sequencing platform. About 79 Gb base pair (583 million reads) of sequencing data were generated and about 831 thousand SNPs were discovered from this panel. We also identified about 17 thousand low-missing data rate (LMD<50) SNPs. The SNPs discovered by tGBS will be combined with SNPs identified from traditional Genotyping-by-Sequencing generated by Dr. Ed Buckler's and Dr. Jianming Yu's groups for future GWAS study to discover genetic controls of sorghum biomass yield-related plant architecture traits. Both diversity panels (photoperiod sensitive and insensitive sorghum lines) were planted at two experimental stations in Ames and Boone, IA (Kelley farm and Agriculture Engineering and Agronomy farm). The experimental design was, as in previous years, a randomized complete block design, in which blocks were split by plant height. This design was implemented to minimize the unfair competition that could be generated by planting two different sorghum types (e.g. grain vs. photoperiod sensitive). Each genotype was planted in a 2-row plot with 90-inch row spacing between plots and 60-inch row spacing between rows of the same plot and two replicates per genotype. The following phenotypic data were collected and statistical analysis performed during this reporting period: Photosynthesis and light-adapted fluorescence parameters have been statistically analyzed to determine Best Linear Unbiased Predictors (BLUPs) of genotypes for further association analysis. Weather data collected at experimental farms have been included in linear models as covariates. Stand counts, to determine the number of plants in each plot, were collected from both locations. These data will be utilized to correct biomass yield for variations in plant density. Images were collected weekly from both locations. Objective 2: Develop a high-throughput phenotyping robot to measure growth rates. We upgraded our imaging system using NIR-converted stereo cameras and collected data from a subset of sorghum varieties at harvest in 2014. We parallelized and improved our current stereo matching algorithm for GPU and developed batching processing framework using MPI on ISU HPC cluster. We also collected NIR stereo images of two locations between mid-June to mid-August in 2015. We added Kinect sensors to our current system and will collect 3D point clouds of a subset of sorghum varieties in September 2015. We calibrated the NIR-converted stereo cameras to rectify image distortion and acquire their relative positions and orientations. We developed sorghum plant segmentation algorithm based on NIR images. We also developed plant localization and stalk width detection algorithm for sorghum at the early stage based on the NIR segmentation images.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2015
Citation:
Bao, Yin; Tang, Lie; Schnable, Patrick S.; and Salas Fernandez, Maria G., "GPU-based Parallelization of a Sub-pixel Highresolution Stereo Matching Algorithm for Highthroughput Biomass Sorghum Phenotyping" (2015). Agricultural and Biosystems Engineering Conference Proceedings and Presentations. Paper 452.
|
Progress 09/01/13 to 08/31/14
Outputs
Target Audience: Genetics researchers and the agricultural community. Changes/Problems: 1. Lodging of tall plants made it difficult for the data acquisition robot to travel between rows. 2. Tall plants sometimes blocked RTK-GPS signal required by the robot. Extending the height of antenna will solve this problem. What opportunities for training and professional development has the project provided? One undergraduate student and an intern received training on the use of LI-COR to be able to collect good quality data and interpret the parameters collected by the equipment. How have the results been disseminated to communities of interest? An oral presentation of the phenotypic data acquisition system and preliminary plant 3D reconstruction results were given at ASABE annual meeting in July 2014 by Tang's group. Salas Fernandez M.G. 2014. Genetic characterization of sorghum biomass yield components over time using field-based high-throughput phenotyping technology. ASA-CSSA-SSSA International Annual meeting, Long Beach, CA. Abstract ID 86748. Invited presentation on November 4, 2014. Salas Fernandez M.G. 2014. Field high-throughput phenotyping technologies for biomass sorghum: current status and future perspectives. R.F. Baker Plant Breeding Symposium, Iowa State University, Ames, IA. March 6, 2014. Salas Fernandez, M.G. Breeding, genomics and high-throughput phenotyping strategies to design a high-yielding biomass sorghum for biofuel production. Energy Biosciences Institute, University of Illinois, Urbana-Champaign, IL. November 2013. Schnable has described the project in over a dozen seminars presented around the world over the last year. What do you plan to do during the next reporting period to accomplish the goals? We are in the process of re-sequencing those sorghum samples that had insufficient numbers of reads. Once this is completed we will re-analyze the SNP data. Photosynthesis and fluorescence parameters will be collected next summer 2015, in one location in Iowa, to expand the phenotypic data for photosynthesis. Both photoperiod sensitive and insensitive panels will be planted in one location to collect images for one last season. We will collect final biomass image data and parallelize the image processing on an ISU HPC (high performance computing) computer cluster. We will develop a robust and accurate stereo matching algorithm for sorghum field scenes. We will also develop a segmentation algorithm to identify sorghum stalks, leaves, and panicles. Based on the segmentation and 3D model, we expect to be able to automatically measure yield component traits such as height, stalk diameter, leaf number, leaf length, panicle size and so on. We will develop a method to quantify sorghum biomass. We also plan to upgrade the phenotypic data acquisition system with IR-converted DSLR (digital single-lens reflex) cameras with onboard memory storage. This will simplify plant segmentation algorithm using normalized difference vegetative index (NDVI) and increase the system travel speed because image transmission and saving to a central computer is the current bottleneck.
Impacts
What was accomplished under these goals?
289 lines of sorghum from the Photoperiod-Sensitive panel were multiplexed and sequenced using tunable Genotyping-by-Sequencing technologies. In total we generated 654 million single-end raw reads from four lanes of an Illumina Hi-seq 2000 sequencer. 457 million were retained after de-barcoding. We discovered 525,911 polymorphic sites. However, the overall missing data was unacceptably high because 180/289 lines of samples did not have a sufficient number of reads (>1 million read/line) for single nucleotide polymorphism (SNP) discovery. Both the photoperiod sensitive and insensitive sorghum panels were planted at two experimental stations in the Ames area, IA (Curtiss farm and Agriculture Engineering and Agronomy farm). Two different designs were implemented this year: a) a plot design in which each line was planted in a 2-row plot with 90-inch row spacing between plots and 60-inch row spacing between rows of the same plot; and b) an individual plant trials in which a single plant of each line was spaced laterally 120 inches. The two different designs were implemented to collect images from very different canopy density situations. The following phenotypic data were collected during the summer of 2014: Photosynthesis and light-adapted fluorescence parameters were collected from 315 sorghum lines of the photoperiod insensitive panel in one of the two 2014 locations. There was significant variation in all parameters, with a positive and significant correlation between photosynthesis and stomatal conductance. Weather data were collected to determine the potential effects of variation of temperature, radiation, wind and humidity on photosynthesis and fluorescence parameters. Plant architecture parameters including plant height, leaf number, leaf angle and stem diameter were manually collected from 30 lines, and these data will be utilized to develop and validate our image analysis algorithms. Stand counts were collected, to determine the number of plants in each plot on both locations. These data will be utilized to correct our biomass yield estimates by variation in plant density. Biomass data for individual plants of the plot design were collected for a subset of 100 lines in one location. The goal was to estimate the proportion of the total aboveground biomass determined by grain. Biomass data at the plot level in both locations utilizing an experimental forage chopper. These data will be utilized to predict biomass yield from digital images and to correlate with growth curves and individual plant architecture parameters. Cameras were calibrated to rectify image distortion and acquire relative positions and orientations. A variety of stereo matching algorithms from Middlebury stereo vision benchmark were tested. Their advantages and limitations for sorghum field scenes were investigated. Some image segmentation algorithms were tested to separate plants from sky and soil. Sky can be robustly removed, while some challenges remain for distinguishing plants from soil. The efficiency of the data acquisition software was improved. The maximum travel speed of the robotic image capture system was improved to 1.2 MPH with all 12 cameras running. Sorghum images were collected from the two locations alternately once per week (as weather permitted) from mid-July to the beginning of September 2014. Each location has a single-plant plot and a 10-feet-range plot. A total about 5,131 images were collected per location per week.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2014
Citation:
Bao, Y., A. D. Nakarmi, L. Tang. 2014. Development of a filed phenotyping robotic system for sorghum biomass yield component traits characterization. ASABE Paper No. 141901199, St. Joseph, MI.
|
Progress 09/01/12 to 08/31/13
Outputs
Target Audience: Genetics research and agricultural community. Changes/Problems: 1. All plants were damaged by hail. Although they recovered to some degree, the phenotypic data are not consistent. Thus we decided to only collect images of different cases to develop the image processing algorithms. 2. Next year the plants will be planted at multiple locations in case of severe weather events. 3. 60" row spacing is too narrow because the leaves are blocking the cameras on the robot. Next year it will be 120". 4. More plant spacing is needed for sample plants since it is extremely difficult to process densely planted sorghums. What opportunities for training and professional development has the project provided? One graduate student was trained on the construction and optimization of the phenotype collecting robot. One undergraduate student and an intern received training on the use of Lic-COR to be able to collect high quality data and interpret the parameters collected by the equipment. How have the results been disseminated to communities of interest? Salas Fernandez, M.G. Improving the photosynthetic capacity of sorghum: current and future perspectives. Great Plains Sorghum Conference and 29th Sorghum Research and Utilization Conference. Kansas State University, August 2012. Schnable, P., Next Generation Genotyping and Crop Improvement. BASF, Research Triangle Park (RTP), NC, February 2013. Schnable, P. Next Generation Phenotyping and Breeding. 6th Annual University of Minnesota Plant Breeding Symposium, Minneapolis, MN, March 2013. What do you plan to do during the next reporting period to accomplish the goals? Photosynthesis and fluorescence parameters will be collected in summer 2014 in two locations in Iowa to complete the phenotypic data needed for association analysis. Both photoperiod sensitive and insensitive panels will be planted in both locations to collect phenotypic data on plant architecture parameters and growth once a week during the summer season. We will develop the image processing algorithms before next growing season. We will improve our current system based on the results from this year and collect phenotypic data for next growing season. We will grow the Yu panel again in 2014, process the data and obtain the measurements of yield component traits throughout the growing season. We will also randomly select samples and take manual measurements for validation. The sorghum lines will be genotyped to validate their identities. Subsequent genome-wide association studies (GWAS) will be conducted using the phenotype data.
Impacts
What was accomplished under these goals?
Seed stocks were increased for the 300 lines in the Yu Panel of photoperiod sensitive lines during the 2012 winter nursery. Two biological replicates of these lines, plus 315 sorghum lines of the photoperiod insensitive panel were grown in IA during the summer of 2013. Leaf samples were collected from the Yu Panel lines grown in the 2013 summer nursery. Freeze-dried leaf samples were powdered, and DNA was extracted. DNA samples are awaiting quality control and subsequent single nucleotide polymorphism (SNP) analyses. Photosynthesis and light-adapted fluorescence parameters were collected from the sorghum lines in the photoperiod insensitive panel. There was a significant variation in all parameters. Preliminary analysis of the data demonstrates that, as expected, there was a positive and significant correlation between photosynthesis and stomatal conductance. Data were collected over 11 days from 10:00am to 5:00pm. Weather data were collected to determine the potential effect of variation on temperature, radiation, wind and humidity on photosynthesis and fluorescence parameters. Preliminary analysis suggests that there was no effect on radiation variation within a day on the photosynthesis. Variation between days will be included in the statistical model to correct for these effects. A phenotypic data acquisition system that automatically collects stereo images of sample plants was constructed. Sorghum images were collected from 4 plant types (single short/young plant, short/young plant with tillers, single tall/mature plant and tall/mature plant with tillers) on which the image processing algorithms will be tested. Biomass yield was determined for a subset of lines (10%) to be used in the algorithm development process and to validate them. Plant architecture parameters including plant height, leaf number, leaf angle, stem diameter, number of tillers, number of internodes, green weight and dry weight were also manually collected. Images were also collected from these for validation and algorithm development. Carbon assimilation through photosynthesis is the basis of crop productivity and it determines biomass yield. When plants are under stress, the excess light energy that cannot be utilized for photosynthesis must be eliminated by photoprotective mechanisms. Significant variation for leaf photosynthetic rates and photoprotection (measured as fluorescence) was identified in sorghum this past summer. This variation will make possible the identification of genes that control these processes and the efficient “design” of germplasm with higher photosynthetic and photoprotective capacity. These materials will produce more biomass under both stress and non-stress environments, could be planted in more marginal areas, and will have yield stability over time and under unpredictable weather conditions.
Publications
|
|